ctx gives AI coding agents project memory. Actx/ folder in your project root with plain markdown files that any agent reads on startup. It knows where the project stands, what patterns work, what mistakes to avoid, and what was in progress when the last session crashed. Loads ~80 lines instead of re-reading everything. Ships 12 dev workflow skills (TDD, debugging, code review, design intelligence). Works on Claude Code, Codex, Cursor, Windsurf, Gemini, Antigravity. Zero dependencies.

Hey, I'm Pulak. I built ctx because I kept losing context. I'm working on a large Laravel + React project (35 models, 40 migrations, 10 modules). Every time I started a new session or switched agents, the AI had no idea what happened before. It would re-read everything, rediscover patterns I'd already established, and sometimes repeat mistakes I'd already fixed. The breaking point was when I had 4 dashboards built across different sessions, each using a different layout. They looked like 4 separate apps. The agent didn't remember that we'd standardized on one layout two sessions ago. So I built a memory layer. Not a database, not a vector store. Just markdown files organized like a book. Table of contents, knowledge pages, session summaries. Any agent on any platform can read them. The crash recovery part came from personal pain too. Session disconnects mid-refactor, you open a new one, and the agent has no idea you were halfway through rewriting 4 files. wip.md solves that. It tracks what's in progress, step by step. If it exists when a new session starts, the previous one was interrupted. ctx is Apache 2.0, owned by ATUL AI. Free for everyone. GitHub: https://github.com/ATUL-Labs/ctx

ctx is on active development; now you can see your live development your work in progress and all, have a look, build your app using even small models, thank me later :).

lex

Keep your AI coding sessions alive.Never lose context again.

Keep your AI coding sessions alive.Never lose context again.